contiguous

→Tensor

contiguous(input: Tensor)

source

Return a tensor with C-contiguous (row-major) memory layout.

If input is already contiguous this is a no-op and returns the same tensor; otherwise a fresh buffer is allocated and the data is copied into row-major order.

Parameters

inputTensor

Possibly non-contiguous source.

Returns

Tensor

Contiguous tensor with identical shape, dtype, device and values.

Notes

Operations that rely on stride assumptions (e.g. view, certain BLAS kernels) require contiguous input. Calling contiguous after a sequence of stride-changing ops (transpose, permute, expand) is the canonical fix when downstream ops complain.

Examples

>>> import lucid
>>> x = lucid.zeros(3, 4).transpose(0, 1)
>>> x.is_contiguous()
False
>>> lucid.contiguous(x).is_contiguous()
True